Looking for a better audio recording interface for my Gateway (Motion
Computing) Tablet PC, I ran across the M-Audio Transit USB. It
immediately worked for playing sound, but I've been through a major
hassle with recording. It turns out the Transit is designed with 2.24
VDC of bias on each analog audio input, to power cheap PC microphones.
When connected to other audio devices, this bias voltage fights the
source you are trying to record. Depending on the source impedance, it
might even draw enough current from the Transit's tiny USB-driven
power supply to interfere with the USB signal and cause full-volume
hissing in place of music. With any audio source, I found I could have
only one end of the USB cable connected to grounded equipment. If the
audio source was grounded, the computer could not be - not even by the
supposedly isolated AC-powered battery charger.

I spent over a month eMailing M-Audio about these issues, and received
no help, only the typical runaround about drivers, SpeedStep, and the
kind of IRQ conflicts that went away with the ISA bus. I finally
started calling their phone support with the single DC bias issue, and
in only a day got them to confirm that the 2.24 V bias was by design.
In another hour I had cut the traces and solved almost all of the
problems with recording through the Transit. Suddenly there was no
more hissing, no more problem with using the battery charger while
recording, and the extreme loading of the source signal went away.

I don't guarantee that I've found the best way to accomplish removing
the DC bias, but here's how I did it. (For a photo of the area I'm
referring to, see http://www.psychoros.org/Transit-Input.jpg.)

The output end of R4 is a 4.4V supply. R33, R31, and C36, and R34,
R32, and C37, form a pair of voltage dividers, creating two
independent filtered sources of the 2.2V bias voltage. R21 and R22
connect the bias voltages to the actual analog audio input lines, and
in the process connect C36/37 between the audio inputs and ground.
(The equivalent sized C32 and C33 connect the audio inputs to the ADC
circuits, so the loading is significant.)

You can probably see my ugly cuts in the traces to the right of R21
and R22, separating the DC bias supplies from the audio inputs, in the
photo. (Guess I need a new X-Acto blade...) Those two tiny cuts solved
the problems with my Transit and the Tablet PC, stopped the excessive
loading of the audio source, and dramatically improved the recorded
sound. With this issue fixed, I'm very pleased with the sound quality
of both recording and playback through the Transit. I can't say the
same for M-Audio's product documentation or support.

Loren